File size: 4,076 Bytes
e73d316 f120be8 c269194 f120be8 c269194 f120be8 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 |
---
title: MCP NLP Analytics
emoji: 📊
colorFrom: indigo
colorTo: blue
sdk: static
app_file: index.html
pinned: false
---
# Sentiment Evolution Tracker – MCP Monitoring Stack
Sentiment Evolution Tracker is an enterprise-ready monitoring stack that runs as a Model Context Protocol (MCP) server. It combines local sentiment analytics, churn prediction, alerting, and reporting, and can operate standalone or alongside Claude Desktop as an intelligent assistant.
## Why This Exists
Traditional "use Claude once and move on" workflows do not keep historical context, trigger alerts, or generate portfolio-level insights. Sentiment Evolution Tracker solves that by providing:
- Automated trend detection (RISING / DECLINING / STABLE)
- Churn probability scoring with configurable thresholds
- Persistent customer histories in SQLite
- Real-time alerts when risk exceeds 70%
- ASCII and HTML visualizations for demos and stakeholders
- Seven MCP tools that Claude (or any MCP-compatible LLM) can invoke on demand
## 🎥 Demo Video
[Watch Demo](data/demo.mp4)
---
## Installation
```powershell
cd mcp-nlp-server
pip install -r requirements.txt
python -m textblob.download_corpora
python -m nltk.downloader punkt averaged_perceptron_tagger
```
## Daily Operations
- `python init_db.py` – rebuilds the database from scratch (reset option)
- `python tools\populate_demo_data.py` – loads deterministic demo customers
- `python tools\dashboard.py` – terminal dashboard (Ctrl+C to exit)
- `python tools\generate_report.py` – creates `data/reporte_clientes.html`
- `python src\mcp_server.py` – launch the MCP server for Claude Desktop
## MCP Tool Suite
| Tool | Purpose |
| --- | --- |
| `analyze_sentiment_evolution` | Calculates sentiment trajectory for a set of messages |
| `detect_risk_signals` | Flags phrases that correlate with churn or dissatisfaction |
| `predict_next_action` | Forecasts CHURN / ESCALATION / RESOLUTION outcomes |
| `get_customer_history` | Retrieves full timeline, sentiment, and alerts for a customer |
| `get_high_risk_customers` | Returns customers whose churn risk is above a threshold |
| `get_database_statistics` | Portfolio-level KPIs (customers, alerts, sentiment mean) |
| `save_analysis` | Persists a custom analysis entry with full metadata |
## Data Model (SQLite)
- `customer_profiles` – customer metadata, lifetime sentiment, churn risk, timestamps
- `conversations` – every analysis entry, trend, predicted action, confidence
- `risk_alerts` – generated alerts with severity, notes, and resolution state
Database files live in `data/sentiment_analysis.db`; scripts automatically create the directory if needed.
## Claude Desktop Integration
`config/claude_desktop_config.json` registers the server:
```json
{
"mcpServers": {
"sentiment-tracker": {
"command": "python",
"args": ["src/mcp_server.py"],
"cwd": "C:/Users/Ruben Reyes/Desktop/MCP_1stHF/mcp-nlp-server"
}
}
}
```
Restart Claude Desktop after editing the file. Once connected, the seven tools above appear automatically and can be invoked using natural language prompts.
## Documentation Map
- `docs/QUICK_START.md` – five-minute functional checklist
- `docs/ARCHITECTURE.md` – diagrams, module responsibilities, data flow
- `docs/HOW_TO_SAVE_ANALYSIS.md` – practical guide for the `save_analysis` tool
- `docs/EXECUTIVE_SUMMARY.md` – executive briefing for stakeholders
- `docs/CHECKLIST_FINAL.md` – submission readiness checklist
## Tech Stack
- Python 3.10+
- MCP SDK 0.1+
- SQLite (standard library)
- TextBlob 0.17.x + NLTK 3.8.x
- Chart.js for optional HTML visualizations
## Status
- ✅ Production-style folder layout
- ✅ Deterministic demo dataset for the hackathon video
- ✅ Comprehensive English documentation
- ✅ Tests for the `save_analysis` workflow (`tests/test_save_analysis.py`)
Run `python tools\dashboard.py` or open the generated HTML report to verify data before your demo, then start the MCP server and launch Claude Desktop to show the agentic workflow in real time.
|